[SR-Users] Segfault in Kamailio 1-3-4 to 1.5.5

marius zbihlei marius.zbihlei at 1and1.ro
Tue Mar 1 11:20:59 CET 2011


On 03/01/2011 11:02 AM, Andrew O. Zhukov wrote:
> On 03/01/2011 10:49 AM, Daniel-Constantin Mierla wrote:
>    
>>
>> On 3/1/11 8:41 AM, Andrew O. Zhukov wrote:
>>      
>>> I someone interested in .
>>> It's the old coredumps from 1.3.4. It's really much stabile then 1.5.5
>>> I did degrade today night
>>>        
>> version 1.3.x is openser only which became later kamailio, practically
>> is no other option for this version.
>>
>> Have you considered upgrading to latest stable (3.1.x) instead of
>> downgrade?
>>      
> Daniel,
> I sent you my config.
> How can I do it on a hi usage production server for a one night.
> The lot of fixes for a different buggy customers SIP and NAT devices
> which is impossible to retest again.
>
> I'll try opensips and possible will be back.
>    

Hello,

I had been checking the coredumps you provided for a while. I don't 
think they are very useful because for me this looks like an Heisenbug. 
The coredumps only show the result of the memory corruption and not the 
cause.

Daniel has asked you for some input when compiling with memory debug on 
(to see if canary values where overwritten by what operations). I have 
not seen this output yet (don't know if you send it privately to him, or 
I have missed it on the list). Try to compile again with debug memory 
support, and set logging to a apropriate level.

An idea is to set a special server with a special version of Kamailio. 
Minimize the number of children (use only one worker) and use just part 
of the traffic so you have a easier debugging. The bug affects private 
memory so the number of children should not be an impact. Also you might 
want to increase the PKG_MEM_SIZE from the default value of 4 MB to 
something bigger (try 10-15 MB or more). See if this has an impact on 
the bug (it might be caused by fragmentation in this case a bigger pool 
might help).

If all else fail, I strongly suggest dropping pkg_malloc all together, 
and using libc's Malloc() instead. This is done at compile time by 
removing the -DPKG_MALLOC from Makefile.defs and recompiling.. This 
should fix your bug

Marius


>> Cheers,
>> Daniel
>>
>>      
>
> _______________________________________________
> SIP Express Router (SER) and Kamailio (OpenSER) - sr-users mailing list
> sr-users at lists.sip-router.org
> http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-users
>
>    




More information about the sr-users mailing list